Repetitive-element PCR (rep-PCR) fingerprinting is a promising molecular typing tool for Escherichia coli, including for discriminating between pathogenic and nonpathogenic clones, but is plagued by irreproducibility. Using the ERIC2 and BOXA1R primers and 15 E. coli strains from the ECOR reference collection (three from each phylogenetic group, as defined by multilocus enzyme electrophoresis [MLEE], including virulence- associated group B2), we rigorously assessed the effect of extremely elevated annealing temperatures on rep-PCR's reproducibility, discriminating power, and ability to reveal MLEE-defined phylogenetic relationships. Modified cycling conditions significantly improved assay reproducibility and discriminating power, allowing fingerprints from different cyclers to be analyzed together with minimal loss of resolution. The correspondence of rep-PCR with MLEE with respect to tree structure and regression analysis of distances was substantially better with modified than with standard cycling conditions. Nonetheless, rep-PCR was only a fair surrogate for MLEE, and when fingerprints from different days were compared, it failed to distinguish between different clones within all-important phylogenetic group B2. These findings indicate that although the performance and phylogenetic fidelity of rep-PCR fingerprinting can be improved substantially with modified assay conditions, even when so improved rep-PCR cannot fully substitute for MLEE as a phylogenetic typing method for pathogenic E. coli.